Team C - Identification of clean housing opportunities

Title: Group C - Energy Use TBC

Mission: “halve the energy use of new buildings by 2030”

Clean Growth

Stakeholder: HM Government: Department for Business, Energy & Industrial Strategy.

Research question

How is your analysis related to your theme? Analysis looked to compare housing distribution,

What is your research question? As part of the industrial Strategy - Clean Growth Grand Challenge there has been a call for the construction industry to reduce whole life cost for assets by 33%, greenhouse gas emission by 50% to drive the UK construction industry as a global leader in clean growth.

Context

#Use tools as you need to. 
#Some in the teams has experience running python with R if you are more comfy with that.

#Ploty is nice to use, but screenshot from excell with do too. 
library(plotly)
#library(circlize)
library(knitr)
library(geojsonio)
library(sp)
library(tmap)

Datasets

Datasets you used:
Building efficiency - EPC ratings - https://www.gov.uk/government/statistical-data-sets/live-tables-on-energy-performance-of-buildings-certificates London Energy Map project - CSE - https://www.cse.org.uk/projects/view/1259#Energy_consumption_data__domestic_ London Postcodes - Doogal - https://www.doogal.co.uk/london_postcodes.php

What data is available to measure progress on your problem?

Tools

R QGIS Python

Analysis

##1. Housing distribution # Clustering of High Energy Consumption - Spatial Statistics

First, we had a look at the spatial distribution of residential electricity consumption across London.

# Add electricity consumption map

The map shows that there appears to be a concentration of high-consuming regions in central London.

The first thought that comes to mind is that this is a function of housing density: more populated areas consume more electricity because there are more houses.

# Add housing density map

library(plotly) # Interactive visualisation
data <- read.csv("./Test.csv")


#saveRDS(object, file = "my_data.rds")
# Restore the object
#readRDS(file = "my_data.rds")

#colnames(data) <- c("colum1name", "column2name")
f <- list(
  family = "Courier New, monospace",
  size = 14,
  color = "#7f7f7f"
)

x <- list(
  title = "Number of Households with at least one user",
  titlefont = f
)
y <- list(
  #nticks = 10,
  #domain = c(0, 90000000),
  #range = c(0,80000000),
  title = "Total domestic electricity consumption (kWh)",
  titlefont = f
  )

scatter <- plot_ly(
  data, x = ~Households_Least_One, y = ~Electricity,
  type="scatter", mode="markers",
  color=~Electricity, size=~Households_Least_One)  %>%
  layout(xaxis = x, yaxis = y)

#layout(yaxis = list(title ="All Households"), xaxis = list(title ="Total domestic electricity consumption (kWh)")

scatter

## Warning: Ignoring 1577 observations

## Warning: `line.width` does not currently support multiple values.

#htmlwidgets::saveWidget(as.widget(p), "test.html")

Checking the housing density (map) and comparing it to the electricity consumption (scatter plot), we find that that three is no relationship i.e. the regions of high electricity consumption are real.

The second thought is whether the apparent clustering of high electricity consumption regions is real, and not just our eyes plating tricks. We defined “high domestic electricity consumption” as > 3.5 million kWh.

Admin

After selecting the regions that fit this criteria, we used Ripley’s K function to find whether (A) we had clustering and (B) the approximate size of the clustering. Care was taken to make sure the data was transformed from Mercator to Lamberts Equal Area Projection.

We tested our distribution against a null hypothesis (random distribution) and found that the data is indeed clustered (i.e. has an inhomogeneous distribution). This shows up well in a kernel density plot of the points.

I then used an L function to examine the radius over which this clustering was significant, which is approximately ~2.5 km.

Admin

The conclusion from this analysis is that houses with high electricity consumption is indeed clustered, and this is not a function of housing density. We then looked to explore other factors that might be determining this clustering.

First, we had a look at the spatial distribution of residential electricity consumption across London.

[Admin](.) # Add electricity consumption map

The map shows that there appears to be a concentration of high-consuming regions in central London.

##3. Fuel poverty v income

#

#

Admin

##4. Income and consumption

# Add electricity consumption map

# Add electricity consumption map

library(plotly) # Interactive visualisation
data <- read.csv("./Fule povery London.csv")

#saveRDS(object, file = "my_data.rds")
# Restore the object
#readRDS(file = "my_data.rds")

#colnames(data) <- c("colum1name", "column2name")
f <- list(
  family = "Courier New, monospace",
  size = 14,
  color = "#7f7f7f"
)

x <- list(
  title = "Number of Households with at least one user",
  titlefont = f
)
y <- list(
  #nticks = 10,
  #domain = c(0, 90000000),
  #range = c(0,80000000),
  title = "Fuelpoor Households",
  titlefont = f
  )

scatter <- plot_ly(
  data, x = ~ households , y = ~fuelpoor_households,
  type="scatter", mode="markers",
  color=~ fuelpoor_households , size=~ households )  %>%
  layout(xaxis = x, yaxis = y)

#layout(yaxis = list(title ="All Households"), xaxis = list(title ="Total domestic electricity consumption (kWh)")

scatter

## Warning: `line.width` does not currently support multiple values.

#htmlwidgets::saveWidget(as.widget(p), "test.html")

# Add electricity consumption map

##5. Occupants v consuption

# Add electricity consumption map

# Add electricity consumption map

##6. Type of ownership v consuption

# Add electricity consumption map

# Add electricity consumption map

##7. Age v consuption

# Add electricity consumption map

# Add electricity consumption map

# Add electricity consumption map

##8.Floor area (sqm) Analysis

Floor Area vs. Current heating costs

Trying to understand what the relationship between dwelling sizes and other factors. - Dataset: Energy usage statistics (gas, electric) - How: Using Python and the API we loaded data and analysed it using python libraries - Example of API request:

url = "https://epc.opendatacommunities.org/api/v1/domestic/search?local-authority=E09000015"
headers = {'Accept': 'application/json', 'Authorization': 'Basic c2lydmFuM3RyQGdtYWlsLmNvbToxYjZjMTFhZGFlNjliZWMyMmU5MzBhOGY4YzllNjdkOTViNTQ0ODll'}
r = requests.get(url, headers=headers)

knitr::include_graphics("img/flrareavsheat.png")

• So what? We see that property internal areas might reveal factors that will impact future house building - e.g. typical dwelling sizes.

Floor area vs. Environmental impact

• Dataset: Energy usage statistics (gas, electric)
• How: Using Python and the API we loaded data and analysed it using python libraries

knitr::include_graphics("img/floorareavsenvimpact.png")

• So what? Environment impact is largely clustered around smaller dwellings - therefore we could focus on the geospatial analysis in the future to see where these properties are? e.g. are they within certain demographics region or are they close? This can impact policy planning such building heat networks.

Suggestion: Do a broadsweep of the situation first.

Source: Caveat to keep in mind: Reported by Bonnie once, August 2019.

Suggestion: Then get specific with spatial analysis and modelling

Your recommendation

Invest in existing building stock - retrofit all existing with smartmeters to allow for more accurate datasets

Ownership influences usage and poverty values -

Further Notes for Discussion

Include any cool stuff you found revelant. Construction sector deal - https://assets.publishing.service.gov.uk/government/uploads/system/uploads/attachment_data/file/731871/construction-sector-deal-print-single.pdf?_ga=2.136256562.2056560497.1566385009-1704179078.1558520267 Industrial Strategy - https://www.gov.uk/government/topical-events/the-uks-industrial-strategy